Toward Capturing the Exposome: Biomarker Variability
and Co-Exposure Patterns in the Shared Environment
Ming Kei (Jake) Chung1, Rajeshwari Sundaram2, Kurunthachalam Kannan3, Germaine Buck Louis4, Chirag J Patel1 1Department of Biomedical Informatics, Harvard University 2Biostatistics and Bioinformatics Branch, The National Institutes of Health 3Division of Environmental Health Sciences, New York State Department of Health 4Office of the Director, Division of Intramural Population Health Research, The National Institutes of Health
INTRODUCTION Figure 2 (Left)
Exposome correlation globe showing the Mining the exposome could allow us to discover the dark matter relationships between females, males and that helps to explain common diseases. However, the exposome couples. Only rs greater than 0.25 and smaller is a dynamic entity that varies from time-to-time (temporal) and than -0.25 were shown as connections in the place-to-place (spatial). While it is not possible to track and diagram. Red line denotes positive correlation assess every subject, unit-level ascertainments of exposures could and dark green line denotes negative one. be a solution. Color intensity and line width are proportional Because family members spend a significant amount of time at to the size of the correlation. Within-class and home and many known sources of exposures are shared among between-class correlations are shown outside them (e.g. food and water), we hypothesize that shared and inside of the track respectively. household could have influential effects 1) on the exposure levels Correlations in couples are indicated by the and 2) co-exposure patterns between the members. lines linking females and males (i.e. crossing the vertical-half of the globe). METHODS CONCLUSIONS Study Leveraged 501 pairs of couples from the Longitudinal Investigation of Fertility and the Although individuals spend a significant Subjects Environment Study (2005 to 2009) amount of time at home, shared environment is not a significant factor explaining co-variations of 128 chemical biomarkers. Used 128 biomarkers from 13 chemical classes Exposures (Figure 2) Co-exposure patterns of the exposome, Most are endocrine disrupting chemicals especially for persistent organic EDCs (EDCs). E.g., polyfluoroalkyl substances (PFASs) (e.g. PCBs), are tightly correlated but similar in males versus females separately. 1) Estimated the % variance explained (adjusted coefficient of determination, R2) by shared household with model residuals IMPLICATIONS Statistical Log(chemical) ~ age + sex + Analyses total_lipids/ creatinine To study the elusive environment: 2) Estimated the Spearmans rank correlation (rs) Lipophilicity and persistency of the with model residuals chemicals are important factors justifying Log(chemical) ~ total_lipids/ creatinine the use of unit-level instead of individual- based exposure assessment. Understanding the dense correlations in RESULTS the exposome could help to increase the power of exposome-wide association study and identify the key disease- associating entities in mixtures of exposures.
TOOL
Figure 3 The R code for the exposome globe can be
easily modified to fit for other larger and Boxplots of rs within different chemical classes. A) complex exposome-wide correlation Females; B) Males; and C) Couples. Certain classes contain analyses. Source code will be publicly only 1 pair of correlation (paracetamols in females, available on GitHub (via MIT license) when paracetamols in males, and cotinine in couples). All we publish our findings. represents the grouping by the correlation of all pairs of chemicals available. Horizontal line drawn across the chemical classes is equal ACKNOWLEDGEMENTS Figure 1 to the 95th percentile of the null distribution obtained from permuting the concentrations of all chemicals. Summary of the percentage variance explained by the shared environment. Boxplots of the adjusted coefficient of determination within different chemical classes are shown. Shared household explained less than 20% of the total variance in 11 of the 13 EDC classes.